Datasets:

ivopersus
/

LeakProTabular

license: cc-by-4.0
language:
  - en
pretty_name: UCI Tabular Benchmark Sample
task_categories:
  - tabular-classification
  - tabular-regression
tags:
  - uci
  - tabular
  - classification
  - binary
  - multiclass
  - regression
size_categories:
  - 10K<n<100K
  - 100K<n<1M

UCI Tabular Benchmark Sample

This repository contains a small collection of tabular datasets mirrored from the UCI Machine Learning Repository and prepared for convenient experimentation.

The datasets are organized by common ML task type:

Binary Classification

Adult dataset

Bank Marketing dataset

Multiclassification

Covertype dataset

Statlog (Shuttle) dataset

Regression

Year Prediction MSD dataset

Source and License

All datasets in this repository are sourced from the UCI Machine Learning Repository and are used under Creative Commons Attribution 4.0 International (CC BY 4.0).

CC BY 4.0 license text: https://creativecommons.org/licenses/by/4.0/

Upstream dataset pages (UCI)

Adult (Census Income): https://archive.ics.uci.edu/dataset/2/adult
Bank Marketing: https://archive.ics.uci.edu/dataset/222/bank+marketing
Covertype: https://archive.ics.uci.edu/dataset/31/covertype
Statlog (Shuttle): https://archive.ics.uci.edu/dataset/148/statlog+shuttle
YearPredictionMSD: https://archive.ics.uci.edu/dataset/203/yearpredictionmsd

Modifications and Data Processing Notes

Some modifications may have been applied for usability and consistency. These modifications are intended to be non-substantive and not to change the meaning of the data.

Typical changes include:

Adding column headers where the original files did not include headers.
Converting original formats (for example space-separated or other delimiters) into .csv.
Normalizing line endings and basic formatting fixes to improve parsing.
In some cases, reorganizing files into a standard folder structure (for example train.csv and test.csv).

Unless explicitly stated in a dataset folder README (if present), no attempt was made to:

Remove rows or features
Alter feature values
Rebalance classes
Impute missing values

If you require a byte-for-byte identical copy of the upstream distribution, please download directly from the corresponding UCI page.

Missing Values

Missing values are preserved as in the upstream sources. Depending on the dataset, missingness may appear as empty fields, ?, or other dataset-specific markers. Refer to each dataset's UCI documentation for the authoritative description.

Intended Use

This pack is intended for:

Tabular model benchmarking (linear models, tree models, neural networks)
Privacy and security research on tabular learning pipelines, including:
- reconstruction and gradient inversion attacks
- defenses such as clipping, noise injection, discretization, constraint-aware decoding
- evaluating reconstructibility using feature-level and record-level metrics

It is not intended for making decisions about individuals or for any high-stakes deployment